OAS Playout Hastings Rock

Trial Log - May 2002

On Air....

As I previously mentioned, I regarded this as the biggest trial my package could go through - so here's a journey through the 28 days of broadcasting in May....

Day -1:

Much of the music was loaded on the system through the last week but clearly the workload to get all the files indexed and loaded into the database had been underestimated - as frequently mentioned by Simon (the guy doing most of it) - "we should have started this lot this time last year". He'd been up half the night throughout the week sorting out what amounted to nearly 4000 tracks.

The lack of all the music loaded also meant the playlist wouldn't be available for the first day. The list of tracks to be played is created as a separate database which is then 'matched' to the central music store from all the physical MP3 tracks. Once the two lists are matched together, the running order is generated automatically from a custom bit of software I wrote and takes only a couple of minutes.

Adding to the general excitement was my ability to trash the existing music database at 8pm that evening. Oops.

off-air recording studio

Day 1:

Midnight and the first test was 6 hours of automation picked pretty much at random from the music we had loaded. Since I'd been heavily running the software for hours on end in automatic mode I wasn't too worried about this - I'm more concerned about what odd combinations the jocks could invoke to break it all. Still, there's a bit of a buzz you get when driving away from the place at half midnight realizing you are in at least part responsible for generating what's going out over the airwaves just now.

in the kitchen, simon rectifies my deletion of one of Sweden's finest...

3am - lot of dead air space. Not good but as it transpires this is down to not enough music being loaded into the player - so future improvement number one - put in a time estimate of the tracks loaded.

6am and the breakfast show starts but we've still not got a playlist so I'm now involved in cataloguing bits of the remaining tracks in an effort to get it all done today. We plan to be up in the studio by lunchtime with a complete database and 7gig of music to load on - whilst they're using the system.

Hampered again by my ability to destroy databases - this time in removing all tracks by one Nils Lofgren which despite being told otherwise am convinced must be Swedish or Norwegian.

Wonder how many hits I'm getting from Nils Lofgren fans now (actually that's Lofgren; Nils on our system). Despite my attempt to delete all mention of him, he kept cropping up the form of duplicate tracks all throughout the afternoon....

By early evening the cataloguing was complete and we left it copying the remaining 3gig of music onto the toaster to head off for a celebratory curry and a few beers - quick plug there to the Tandoori Ghor, St. Leonards sea front - fine food as always.

We rolled back up to the pub around 10pm to check on the progress and hit the magic button to roll out the playlist ready for first thing tomorrow.

Week 1:

Completion of the first day without any on-air incidents was a relief and the first week proved to be a continuation of that trend - many of the comments I've had have been favorable to say the least and no-one that I've been aware of has had any significant problems with it - ease of use seems to be the watchword.

A minor software change went in over the first few days - principally to allow reversion to the main playlist after any other selection had been made.

Behind the scenes a few technical glitches cropped up - Simon spotted a deficit of 1000 odd tracks not in the running order on day 2 which was a bit of an oversight by me - turns out an earlier change had been 'undone' since I'd last tried a playlist generation but a bit of donkeywork later and the remaining tracks were added.

One thing being showed up at this time was how disjointed the loading of new material onto the system is - whilst it is a fairly straightforward import process the fact that it involves switching between several packages, running up Windows Explorer to copy a few files around ends up making it too complex. Not surprising really since it was only a rough and ready interface but it's confirmed my belief that I need a decent interface to go on before making the package commercially available.

With network hub and server located away from the two studios, it's dead easy to shove a laptop or other PC into the system to make minor tweaks without causing disruption...

toaster (the music server) runs the backup script overnight to copy any new files to the removable drive in the main studio - it emails the results to me nightly so I can check all is working well.

Week 2-3:

Nick - at the controls and thus far failing to break anything...

All continues to go startlingly well, again a distinct absence of any technical hitches which might upset the normal running of things. On the Sunday at the end of the first week the entire music database managed to be loaded into the jingles database - again highlighting the need for a better front end on this part of the system but this was fairly easily rectified by a mass record delete. Something I didn't realize would be happening was the chopping and changing of different ads - the current database doesn't offer an easy way of deleting stuff since it involves not only the records being deleted but the physical MP3 files as well.

The feedback I was getting continued to be good as well - some suggestions for minor layout changes which I will incorporate into future versions including a history of recent tracks played. One thing that was being mentioned was a delay being incurred between the Start button being pressed and the track physically starting. I'm aware this can happen (see the Beta development page) but this is probably exaggerated by gaps at the start of the actual MP3 music files. If a CD track has a second of silence at the start then so will a resultant MP3. To get round this, OAS Playout supports Cue Start points but they'd not been set on any of the Hastings Rock tracks because of the time consuming process of setting them manually. I had a play and came up with a bit of code that would do a scan of the tracks and automatically put an 'optimal' cue point in for all the tracks. A week after putting this change in I'm hoping it's improved things somewhat.

Week 4: Conclusions

The Playlist looped around for the first time at the start of the week, showing up that towards the end of the list some tracks were repeating quite frequently - I knew the list got a bit 'ropey' towards the end as the number of tracks available for play gets fewer and fewer but I suspect this is due in part to it being stitched together in two parts early on due to a bug in the list generation program. Next time round this shouldn't happen or at least will be much reduced.

The toaster also stopped sending the backup log emails - looks very much like a fault with the modem or PC hardware there.

they stuck this on the Queensway bridge - at 9:30pm, in the dark....

From an operational and reliability standpoint it went really well - I'm still amazed we didn't get one lock up or crash. The assortment of people I spoke to who used the playout package seemed happy with it after an initial period of settling down, the flexibility of being able to play everything from any of the players seemed to be a key thing - everyone adopted there own style of using it. It worked "as advertised on the tin" someone said.

3 key things came out of it for the future:

Unsurprisingly the database front end needs doing - even though everyone eventually got to grips with it, there's still far too much scope for messing things up - but then I knew this anyway.
Spurious 1 second delays occurring when you hit the Start button, whilst significantly improved by the injection of the cue points mid way through the broadcast period, this is an issue I'm aware of (see the Development Status page) as a problem with one of the third party components I use with the Player. Unfortunately this causes people to start pre-emptying the system - since it doesn't appear to have the response time of a cued CD, they tend to hit the Start button about a second before they want the track to begin which I see as a step backwards from using CDs.
The output quality isn't up to CD standard, it's got a 'flat' sound to it - something I wasn't aware of but it was mentioned by their main engineering guy who know's what he's talking about. We were using pretty good quality sound cards (a SoundBlaster Live and SoundBlaster AWE64 Gold) and he said that if you compare the output of Winamp with a CD playing the same track there is no perceivable difference. When using the Playout package though it is noticeably not as good.

This last issue is again down to the very same third party audio component which causes the spurious delays and based on the somewhat hefty license fees which would apply when purchasing this for distribution (there's more on this on my Development page) a key thing will be to look at better alternatives to this.

Thoughts, comments or suggestions? feel free to email at the address below....

Home Page for OAS Playout - download a free evaluation copy.

*******************************************************************************************************